Formats raw company context fields into optimized LLM-readable rules and protocols
Formats raw company context fields into optimized LLM-readable rules and protocols
npx skills add https://github.com/mazenkhaledZC/grad-project --skill format-company-context复制此命令并粘贴到 Claude Code 中以安装该技能
Determines if two versions of a company rule are semantically equivalent
Detects escalations, conflicts, and evaluates how the agent handled them (15% weight)
Rates how well the agent acknowledged and responded to customer emotions (10% weight)
Rates accuracy of information provided by the agent against company facts (25% weight)
Rates whether the customer's problem was actually resolved (5% weight)
Classifies overall call severity as Minor/Moderate/Major/Critical (5% weight)
| name | format-company-context |
| description | Formats raw company context fields into optimized LLM-readable rules and protocols |
| version | 1.0.0 |
| author | Team |
Takes raw company context fields (as filled by the QA head) and reformats them into clear, unambiguous, LLM-optimized text. This ensures the model understands the rules precisely regardless of how the QA head worded them.
JSON with the raw context fields from the QA head's input form.
JSON with reformatted fields optimized for LLM consumption:
{
"formatted_rules": [
{
"field": "greeting_script",
"original": "...",
"formatted": "RULE: Agent MUST begin every call with: '...'",
"keywords": ["greeting", "opening"]
}
],
"validation_notes": "Any issues found in the input"
}